Context Window 57
Publishing seems to have been slowing down for the holidays for the last couple of weeks, and while AI isn’t on quite the same trajectory, there are fewer notable developments this week. I hope that you have a good festive break if you’re taking one—I will be, so there will be no newsletter next week. Normal service resumes in January.
Ten weeks ago I wrote about Amazon’s Ask This Book and Story So Far features for Kindle, and the implications of training on publisher content. Where one tech platform goes, others follow: Google is following suit with a Gemini integration in the Google Play Books and Audiobooks app. This kind of integrated AI feels like it will be table stakes for reading systems in 2026. The bigger question for authors and publishers is the training and access model.
OpenAI has upgraded its image generation models, perhaps in response to the success of Google’s Nano Banana Pro. The updated model is significantly better in many respects—in particular, it’s less prone to incorrect text in images, though it’s interesting that the release notes highlight areas where the model has got worse. It’s hard to imagine OpenAI shipping that if they weren’t under such competitive pressure from Google.
Here in the UK, the Department for Science, Innovation and Technology released a statement of progress on its consultation on AI and Copyright. The headline figure, widely reported, was that only 3% of survey respondents agreed with the government’s preferred option of a data mining exception with rights reservation mechanism (an old saying about turkeys and Christmas comes to mind). The government’s full report and economic impact assessments are due by March next year: while the statement acknowledges the importance of the creative industries, it also concludes by stating that, “AI offers the most powerful lever we have for national renewal; it sits at the heart of the government’s plan to kickstart an era of economic growth.”
The UK government also announced a memorandum of understanding with Google DeepMind on use of AI in the public sector: the section that caught my eye is that Google will develop a version of Gemini, grounded in the National Curriculum, for schools. What data and content this is trained on, and the effect it has on traditional educational resources, are key questions for publishers in that sector.
Anthropic released an update to Claude Skills, its feature allowing users to create repeatable workflows within Claude. For organisations on Team and Enterprise plans, this now allows admins to manage and distribute skills to all of their users, so a publisher could create standard, centrally-managed workflows for common tasks.
Mistral shipped a new version of its Optical Character Recognition (OCR) model which performs better at digitising complex documents, forms and tables. For anyone converting printed books or other documents into digital formats, this specialist model looks worth a look.
Few, if any, publishers need to deal with handwriting as part of their content workflows. But this article by a history professor on using AI to digitise handwritten documents is worth reading, partly because it discusses example prompts and gives a really clear ‘how to’ guide, and partly because of its finding that Gemini 3 is delivering results on a par with expert human beings. You probably don’t need to digitise cursive in your work. But I bet there’s a task of equivalent difficulty that you do now where AI’s performance is getting better.
I really enjoyed this opinion piece by Jay Flynn, GM for Research & Learning at John Wiley, on how publishers can innovate around AI, and ended up underlining and highlighting large sections of it. The focus is STM publishing, but I think it would be highly relevant to other publishing sectors as well. I’ll end this year with my three favourite quotes from it, and three questions for you to ask about your business and AI.
“We got so good at protecting what we had that we stopped innovating. Over the last few decades, from an end-user perspective, publishing technology basically sat and watched as a generation’s worth of innovation happened all around it.” What have you shipped this year? What will you ship next quarter?
“If your content can’t be discovered through AI agent search, you’re invisible to a growing portion of users.” Is obscurity a greater threat than piracy?
“It is our responsibility to ensure that AI tools surface research over rumor, elevate evidence over opinion, and champion credibility over confusion, so that trusted knowledge lights the way in an age of misinformation.”
If you don’t make your trusted content easy for AI systems to surface, who—or what—will fill the gap instead?
Happy Holidays.
This was originally published in my email newsletter. To receive weekly updates on how AI is affecting the publishing industry, sign up here.